The effects of thiazolidinediones on blood pressure levels – A systematic review

Abstract

Insulin resistance has been proposed to be the underlying disorder of the so‐called metabolic or insulin resistance syndrome, which represents the clustering in the same individual of several cardiovascular risk factors, such as type 2 diabetes mellitus, hypertension, abdominal obesity, elevated triglycerides and low high‐density lipoprotein‐cholesterol. As far as the connection of insulin resistance and compensatory hyperinsulinaemia with hypertension is concerned, a number of mechanisms possibly linking these disturbances have been described, such as activation of sympathetic nervous system, enhancement of renal sodium reabsorption, or impairment of endothelium‐dependent vasodilatation. Thiazolidinediones (TZDs) constitute a class of oral antihyperglycaemic agents that act by decreasing insulin resistance, and apart from their action on glycaemic control, they have been also reported to exert beneficial effects on other parameters of the metabolic syndrome. In particular, during recent years a considerable number of animal and human studies have shown that the use of TZDs was associated with usually small but significant reductions of blood pressure (BP) levels. Since a possible beneficial action of these compounds on BP could be of particular value for patients with the metabolic syndrome, this review aimed to summarize and evaluate the literature data in the field, derived either from studies that just examined BP levels among other parameters or from studies that were specifically designed to determine the effect of a TZD on BP.

• Blood pressure
• hypertension
• pioglitazone
• rosiglitazone
• thiazolidinediones
• troglitazone

Introduction

Insulin resistance (IR), type 2 diabetes mellitus (DM), hypertension, abdominal obesity, elevated triglycerides and low high‐density lipoprotein (HDL)‐cholesterol are some of the disorders that tend to cluster in the same individual, forming the so‐called “metabolic syndrome” or “insulin resistance syndrome” [1,2]. Some studies show that IR and compensatory hyperinsulinaemia are associated with a higher risk of cardiovascular disease (CVD) independently of the other components of the syndrome [3–6], a finding related to the hypothesis that the clustering of these disorders add cardiovascular risk above the individual risk factors [7,8].

In the initial descriptions of the syndrome, IR was proposed to be the fundamental disorder of it, causally related to the other disturbances through various mechanisms [9,10]. As far as the particular association between IR and hypertension is concerned, several pathways linking IR and compensatory hyperinsulinaemia to blood pressure (BP) elevation have been described. Such pathways are for example the preservation of insulin properties to stimulate the sympathetic nervous system [11–14] and to enhance renal sodium reabsorption [15–17], along with the serious impairment of the ability of the hormone to cause endothelium‐dependent vasodilatation in IR states [18–21], which does not allow vasodilatation to compensate for the former pressor effects [22,23]. Additional mechanisms through which hyperinsulinaemia in IR states could elevate BP include the modulation of transmembrane cation transport and intracellular ion content which leads to increase in intracellular calcium and vasoconstriction [24–26], and trophic actions on vascular smooth muscle cells (VSMC) through the mitogen‐activated protein kinase (MAPK) pathway [27–29].

The thiazolidinediones (TZDs) or glitazones (troglitazone, pioglitazone and rosiglitazone) represent a new class of oral antihyperglycaemic agents that improve tissue sensitivity to insulin and thus decrease IR [30–33]. These compounds act through binding to a nuclear receptor, the peroxisome proliferator‐activated receptor gamma (PPARγ), which regulates the expression of various genes that are involved in adipocyte differentiation and carbohydrate and lipid metabolism [31],[34]. In a considerable number of animal and human studies, TZDs have shown important glucose‐lowering properties and therefore they are currently used as antidiabetic drugs in patients with type 2 DM [30–33]. Accumulating data also indicate that TZDs present beneficial effects on other components of the metabolic syndrome. In particular, in various studies TZDs have been reported to decrease triglycerides, raise HDL‐cholesterol levels and increase the size of low‐density lipoprotein particles. Moreover, they have been found to redistribute body fat away from the central compartment, and to decrease urinary albumin excretion (UAE), plasma C‐reactive protein and plasminogen activator inhibitor‐1 (PAI‐1) levels [30–33].

In addition, a considerable number of animal and human studies have evaluated the effects of TZDs on BP, most of which reported a positive impact of these compounds also on this parameter. In view of the importance that such an action could have for patients with multiple CVD risk factors, the potential effects of TZDs on BP are by no means trivial. The aim of this systematic review was to summarize in a chronological order the to‐date literature data on the effects of troglitazone, pioglitazone and rosiglitazone on BP in an attempt to provide a clarification of this issue.

Methods

A literature search of MEDLINE/PubMed database was performed to identify English‐language original articles in animals or humans published from 1966 until November 2005 that reported data on the effects of the TZDs troglitazone, pioglitazone and rosiglitazone on BP. Search terms used were thiazolidinediones, glitazones, thiazolidinedione, glitazone, PPARγ, troglitazone, pioglitazone, rosiglitazone, BP and hypertension. Reference lists of identified articles were also evaluated for additional relevant papers and information. Moreover, Supplements to Journals included in MEDLINE/PubMed that represent Abstract Books of 2005 international congresses in the fields of Hypertension and Diabetes were searched to identify abstracts with relevant data. Original studies with the above TZDs that either examine BP levels among other parameters or aim specifically to determine their effect on BP were included.

Studies with troglitazone

The prototype of the TZD class of compounds, ciglitazone, was initially reported to reduce BP in obese Zucker rats [35]. As far as troglitazone, the first TZD in clinical use, is concerned, several studies have shown a BP lowering effect in animal models of insulin resistance and/or hypertension. In particular, troglitazone has been found to protect against dietary‐induced BP increase or to reduce BP in obese Zucker rats [36,37], fructose‐fed Wistar rats [38], Watanabe heritable hyperlipidaemic rabbits [39], Otsuka Long–Evans Tokushima fatty rats [40,41], rats receiving a high‐fat diet [42], heminephrectomized Wistar fatty rats [43], fructose‐fed Sprague–Dawley rats [44], in the remnant kidney model of spontaneously hypertensive rats (SHR) [45], and in fructose‐fed borderline hypertensive rats [46]. In a few studies, in streptozotokin‐induced diabetic rats [47–49] and in Sprague–Dawley rats with 5/6 nephrectomy [50], the administration of troglitazone did not affect BP levels, but all these studies aimed primarily to evaluate the effect of the drug on UAE.

The first study to evaluate the effects of troglitazone on BP in humans was that of Kosaka et al., in which 28 Japanese subjects with type 2 DM treated with diet alone and mild hypertension were randomized to troglitazone or placebo [51]. After 12 weeks of treatment, troglitazone was associated with a significant fall of 16 mmHg in office systolic BP (SBP), along with a non‐significant reduction of 3 mmHg in diastolic BP (DBP), as shown in Table I. In the second study, Nolan et al. [52] randomly assigned 18 non‐diabetic obese subjects without hypertension, nine of which had impaired glucose tolerance (IGT), in 200 mg of troglitazone twice daily or placebo for 12 weeks. At the end of the study, a significant reduction of 5±2/4±2 (SBP/DBP) mmHg in ambulatory BP was observed in the troglitazone group, along with a significant increase of about 20% in insulin sensitivity (IS) measured with the euglycaemic–hyperinsulinaemic clamp technique and a decrease of 48% in fasting plasma insulin levels, whereas no changes were observed in the placebo group.

Table I. Studies on subjects with components of the metabolic syndrome reporting effects of troglitazone on blood pressure levels.

Study (Ref.)	Type of subjects	n	Regimens compared	Duration	BP measurement	Mean effect on SBP/DBP vs baseline in Tro groups (mmHg)	IS or IR measurement
Kosaka K et al., J Clin Ther Med 1993 [51]	DM2, hyp	28	400 mg Tro vs plb	12 weeks	Office Bp	−16^a/−3	–
Nolan JJ et al., N Engl J Med 1994 [52]	Obese	18	400 mg Tro vs plb	12 weeks	Ambulatory BP	−5^a/−4^a	e‐h clamp
Ogihara T et al., Am J Hypertens 1995 [53]	DM2, hyp	18	400 mg Tro	8 weeks	Office BP	−9^a/−12^a	–
Ghazzi MN et al., Diabetes 1997 [54]	DM2	114	800 mg Tro vs Gly	48 weeks	Office BP	−1.9/−6.5^a	–
Tack CJ et al., Diabetologia 1998 [55]	Obese	15	400 mg Tro vs plb	8 weeks	Ambulatory BP	−0.8/−3.0^a	e‐h clamp
Imano E et al., Diabetes Care 1998 [60]	DM2	30	400 mg Tro vs 500 mg Met	12 weeks	Office BP	−3/0	–
Sung BH et al., Hypertension 1999 [56]	DM2	22	400 mg Tro vs 20 mg Gly	6 months	Office BP	−9^a/−6^a	–
Kawai T et al., Metabolism 1999 [57]	DM2	48	Diet+400mg Tro vs Sul+400mg Tro vs Diet+Diet	3 months	Office BP	−8^ba	–
Kawano Y et al., J Hypertens 2000 [59]	Overweight, hyp	30	400 mg Tro vs diet	8 weeks	Office BP, ambulatory BP	office: 0.8/0.8 ambulatory: 1.5/1.0	HOMA‐IR
Nakamura T et al., Diab Res Clin Pract 2001 [58]	Obese	29	200 or 400 mg Tro vs plb	12 weeks	office BP	−5^c/−1^c	–
Nakamura T et al., Diabet Med 2001 [61]	DM2	32	400 mg vs 5 mg Gli	12 months	Office SBP	−6^d
Watanabe K et al., J Hypertens 2004 [64]	Hyp	34	AT+400 mg Tro vs AT	6 months	Office BP	−7.4^a/−4.8^a	SSPG

SBP, systolic blood pressure; DBP, diastolic blood pressure; IS, insulin sensitivity; IR, insulin resistance; DM2, subjects with type 2 diabetes mellitus; hyp, subjects with hypertension; Tro, troglitazone; plb, placebo; Gly, glyburide; Met, metformin; Sul, Sulphonylurea; Gli, Glibenclamide; AT, antihypertensive treatment; e‐h clamp, euglycaemic–hyperinsulinaemic clamp; HOMA‐IR, homeostasis model assessment – insulin resistance index; SSPG, steady‐state plasma glucose. ^aSignificant change vs baseline levels or the other group compared. ^bChange of mean blood pressure in the Diet+Tro group vs baseline. ^cCombined data from the two troglitazone groups. ^dMean change for SBP vs baseline in both the groups of patients with micro‐ and macroalbuminuria.

In another Japanese study in 18 hypertensive patients with mild DM, 200 mg of troglitazone twice daily significantly reduced office BP from 164 (SD:±3)/94±2 to 155±6/82±3 mmHg after 8 weeks of treatment [53]. Fasting plasma glucose and insulin levels were also decreased and the reduction in mean BP was strongly correlated with the decrease in fasting plasma insulin levels (r = −0.59, p<0.05). In a multicentre double‐blind study in 114 patients with type 2 DM aiming to compare the effect of troglitazone 800 mg daily to that of glyburide on cardiac function, after 48 weeks troglitazone was associated with a significant reduction of 6.5 mmHg in office DBP vs baseline, in contrast to glyburide [54]. Subsequently, Tack et al. evaluated endothelial dysfunction associated with IR and found no difference in vascular responses to acetylocholine, sodium nitroprusside and L‐NMMA in obese insulin‐resistant subjects compared with lean controls. Therefore, no improvement in the endothelial function was observed after treatment of the obese subjects with 400 mg of troglitazone for 8 weeks. However, troglitazone was associated with a significant decrease of about 3 mmHg in ambulatory DBP, whereas SBP did not change [55].

Sung et al. [56] randomized 22 type 2 diabetic patients to 400 mg troglitazone or 20 mg of glyburide daily to determine the effects of troglitazone in BP at rest and during a mental arithmetic test. After 6 months of treatment, both drugs produced similar reductions in fasting glucose levels but only troglitazone reduced insulin and C‐peptide. In addition, troglitazone was associated with a significant reduction of 9/6 mmHg in office BP at rest vs baseline and a significant decrease of 11 mmHg in SBP response during the mental arithmetic test, whereas glyburide did not affect BP in either case. In a study aiming to evaluate the effect of troglitazone on fat distribution, Kawai et al. added 400 mg of troglitazone per day to type 2 diabetic patients who were treated only with diet therapy or a sulphonylurea. After 3 months of treatment, the addition of troglitazone was associated with a significant reduction of 8 mmHg in mean BP (MBP) in the group of subjects on diet therapy and showed a downward trend on those originally receiving sulphonylurea [57]. In another similar study, 29 subjects with visceral fat accumulation and at least two CVD risk factors including glucose intolerance, hyperlipidaemia and hypertension were assigned to receive either 200 or 400 mg of troglitazone daily or placebo for 12 weeks [58]. After 12 weeks, the observed reductions of office BP in the troglitazone groups were not significant, but in a pooled troglitazone group including only subjects with hypertension SBP showed a significant fall of about 9 mmHg vs baseline.

These beneficial effects of troglitazone were not confirmed in every similar studies. Kawano et al. randomized 30 overweight hypertensive Japanese patients to a weight reduction program based on low‐energy diet or treatment with 400 mg of troglitazone daily [59]. The two groups presented comparable decreases in fasting glucose, fasting insulin and HOMA‐IR (Homeostasis model assessment – insulin resistance) index but only the diet group presented significant reductions in casual or ambulatory BP. However, the duration of this study, as in the case of the study by Tack et al. [55], was rather short (8 weeks), possibly not enabling troglitazone to exert its full insulin‐sensitizing or antihypertensive properties. Two other Japanese studies compared the effect of troglitazone in comparison to metformin [60] and glibenclamide [61] respectively on UAE in type 2 diabetics with microalbuminuria [60] or both micro‐ and macroalbuminuria [61]. In both these studies, troglitazone reduced UAE in microalbuminuric patients, but had no significant effect on BP levels from baseline to the end of the treatment period. However, BP was inadequately recorded in the second study (only SBP was measured) [61]. Moreover, in both studies [60,61], BP levels showed a trend towards reduction in the troglitazone groups and towards increase in all the other groups, but no between‐groups comparisons to detect significant differences between the treatments compared were reported.

The effect of troglitazone on BP could not be evaluated from larger clinical studies as troglitazone was associated with severe hepatotoxicity and rare cases of liver failure and death [62,63]. It was thus withdrawn from clinical use in early 2000. Since then, the only study reporting an effect of troglitazone on BP is again a small intervention study [64], in which 34 patients with mild essential hypertension were treated for 6 months either with antihypertensive drugs or with antihypertensive drugs plus 400 mg of troglitazone. In the troglitazone‐treated group, clinic BP showed a significant mean fall of 7.4/4.8 mmHg, whereas in the other group BP remained unchanged.

Studies with pioglitazone

Pioglitazone was also reported to have beneficial effects on BP levels in studies in insulin‐resistant and/or hypertensive animal models. Until recently, this TZD compound was found to lower BP or attenuate the development of hypertension in Dahl salt‐sensitive and one‐kidney, one clip Sprague–Dawley rats [65,66], obese rhesus monkeys [67], fructose‐ or chow‐fed rats [68], rats receiving a high‐fat diet [69,70], sucrose‐fed or normal‐fed SHR [71–75], genetically obese Wistar fatty rats [76], fructose‐fed Wistar–Kyoto rats [77], fructose‐fed Sprague–Dawley rats [78], stroke‐prone SHR [79,80], Sprague–Dawley rats receiving angiotensin II [81] and streptozotocin‐induced diabetic Sprague–Dawley rats [82]. Again, these significant beneficial effects of pioglitazone on BP were not apparent in a few other animal studies [49],[75],[83,84], several of which [49],[84] primarily aimed to determine the effect of this compound on UAE.

The first human study reporting data on the effect of pioglitazone on BP [85] was published almost a decade after the first relevant study with troglitazone [51]. In that study, Hirose et al. aimed to evaluate the effects of pioglitazone on body fat distribution and various other clinical parameters and added the drug for 3 months in 10 male patients, who were treated with diet alone or with a sulphonylurea. Apart from improving glycaemic control and serum adiponectin levels, at the end of the treatment period pioglitazone was associated with a significant reduction of 11/6 mmHg in office vs baseline BP [85], as shown in Table II. Another relevant study aimed to evaluate the effects of pioglitazone on IS, glycaemic control, lipid profile and office BP in 48 hypertensive but not diabetic patients [86]. After 12 weeks of treatment, pioglitazone produced a greater increase of IS and a greater fall in fasting glucose and insulin levels vs placebo. Office SBP showed a mean decrease of 6.7±15.6 (mean±SD) mmHg in the pioglitazone and of 3.1±16.4 in the control group, but the difference between the groups was not significant. In contrast, DBP showed a significantly greater reduction following pioglitazone treatment (7.9±8.0 mmHg) than with placebo (1.8±8.7 mmHg). It seems that the absence of significance for SBP could be possibly due to the relatively high standard deviations observed, something that should not have occurred if a more precise BP measurement method had been used instead of office BP measurement.

Table II. Studies on subjects with components of the metabolic syndrome reporting effects of pioglitazone on blood pressure levels.

Study	Type of subjects	n	Regimens compared	Duration	BP measurement	Mean effect on SBP/DBP vs baseline in Pio groups (mmHg)	IS or IR measurement
Hirose H et al., Metabolism 2002 [85]	DM2	10	30 mg Pio	3 months	Office BP	−11^a/−6^a	–
Füllert S et al., J Clin Endocrinol Metab 2002 [86]	hyp	60	45 mg Pio vs plb	16 weeks	Office BP	−6.7/−7.9^a	e‐h clamp, HOMA‐IR
Gerber P et al., Curr Med Res Opin 2003 [87]	DM2, (hyp)^b	234	30 mg Pio vs 30 to 45 mg Pio vs 45 mg Pio	20 weeks	Office BP	−10^ac/−8^ac	–
Negro R et al., Minerva Endocr 2003 [88]	DM2, non‐dippers	40	30 mg Pio vs plb	8 weeks	Ambulatory BP	Daytime: −2.6/−2.8 night‐time: −5.3^a/−5.6^a	–
Nakamura T et al., J Diabetes Complications 2000 [92]	DM2	45	30mg Pio vs 5 mg Gli vs 0.6 mg Vog	3 months	Office BP	−6/−4	–
Nakamura T et al., Metabolism 2001 [93]	DM2	28	30mg Pio vs plb	6 months	Office SBP	−4^d	–
Satoh N et al., Diabetes Care 2003 [94]	DM2, (hyp)^b	136	30 mg Pio vs no additional treatment	3 months	Office BP	1/0	–
Belcher G et al., Int J Clin Pract 2004 [89]	DM2, (hyp)^b	3713	Pio vs Met or Gli^e	12 months	Office BP	−1.6^a/−1.4^a	–
Horio T et al., Am J Hypertens 2005 [90]	hyp	30	30 mg Pio	6 months	Office BP	0/−2^a	SSPG
Agarwal R et al., Kidney Int 2005 [95]	DM2, (hyp)^b	44	Pio vs Glip	4 months	Ambulatory BP	3.7/2.2	–
Dormandy J et al., Lancet 2005 [96]	DM2, (hyp)^b	5238	Pio vs plb	34.5 months	Office BP	−3^af/−2^f
Basu A et al., Diabetologia 2005 [Abstract] [91]	DM2	21	45 mg Pio vs 10 mg Glip	12 weeks	Office BP	−8.4^ag	–

SBP, systolic blood pressure; DBP, diastolic blood pressure; IS, insulin sensitivity; IR, insulin resistance; DM2, subjects with type 2 diabetes mellitus; hyp, subjects with hypertension; Pio. pioglitazone; plb, placebo; Met, metformin; Gli, Glibenclamide; Vog, Voglibose; Glip, Glipizide; e‐h clamp, euglycaemic–hyperinsulinaemic clamp; HOMA‐IR, homeostasis model assessment – insulin resistance index; SSPG, steady‐state plasma glucose. ^aSignificant change vs baseline levels or the other group compared. ^bA percentage of the total population was also hypertensive. ^cCombined data for hypertensive subjects from all the pioglitazone groups. ^dMean change for SBP vs baseline. ^eCombined data from 4 randomized, double‐blind studies comparing treatment with pioglitazone, metformin or gliclazide, either as monotherapy or as add‐on treatment on previous metformin or sulphonylurea therapy. ^fMedian change vs baseline. ^gMean change for DBP vs baseline.

In the third study, Gerber et al. [87] randomly added three different schemes of pioglitazone (30 mg for 20 weeks, 30 mg for 12 weeks followed by 45 mg for 8 weeks, or 45 mg for 20 weeks) to the pre‐existing treatment of 234 type 2 diabetic patients. Significant decreases in fasting plasma glucose and HbA_1c from baseline to the end of the study were observed in all the groups, but the decreases of office BP was rather small (up to 3/2 mmHg) and not significant. However, in a group combining hypertensive patients (those with office SBP>140 or DBP>90 mmHg or with background antihypertensive medication) of the total population a significant reduction of 10/8 mmHg vs baseline was observed. As in the two previous studies [85,86], this study also aimed to evaluate the efficacy and safety of pioglitazone in general and not specifically for BP changes [87] and therefore neither ambulatory BP nor IS measurements were performed. On the other hand, Negro et al. [88] randomly added 30 mg of pioglitazone or placebo in the metformin treatment of 40 non‐dipper type 2 diabetic patients in order to study the effect of pioglitazone in night‐time BP using ambulatory BP measurements (ABPM). After 8 weeks of treatment, small and non‐significant decreases in fasting plasma glucose, insulin and HbA_1c levels were observed from baseline to the end of the study in the pioglitazone group and diurnal BP showed a non‐significant drop of about 2.6/2.8 mmHg. However, nocturnal BP showed a significant reduction of about 5.3/5.6 mmHg [88].

Belcher et al. reported combined data from four randomized, double‐blind trials, comparing treatment with pioglitazone, metformin or a sulphonylurea, either as monotherapy or as add‐on treatment on previous metformin or sulphonylurea therapy. In these trials, BP was not a primary endpoint and was measured only as a safety variable in the basis of routine medical care. All treatments produced small decreases in mean office BP levels relative to baseline with pioglitazone having the largest and gliclazide the smallest effect (−1.6/−1.4, −1.5/−1.2 and −0.7/−0.6 for pioglitazone, metformin and gliclazide respectively) [89]. In a more recent study, Horio et al. added 30 mg of pioglitazone daily to the antihypertensive or lipid‐lowering treatment of 30 patients with essential hypertension for 6 months, in order to evaluate the effects of the drug on left ventricular function. Apart from improvement in left ventricular diastolic function, at the end of the study office DBP showed a significant fall of 2 mmHg, whereas SBP did not change [90]. Finally, in a study aiming primarily to evaluate the effects of pioglitazone on body water and body composition, 21 subjects with type 2 diabetes were randomized to receive 45 mg of pioglitazone or 10 mg of glipizide. After 12 weeks of treatment, pioglitazone, in parallel to an increase of body water, was associated with almost significant reductions in office DBP (8.4±4 mmHg, p = 0.05) and mean BP (9.5±5 mmHg, p = 0.08) and a significant decrease in systemic vascular resistance, while glipizide was not. It seems very possible that the absence of significance for BP changes was due to the statistical low power of the study [91].

In contrast to these findings, other studies reported no significant changes in BP after pioglitazone treatment. For example, Nakamura et al. randomized 45 patients with type 2 diabetes and microalbuminuria to 30 mg of pioglitazone, 5 mg of glibenclamide or 0.6 mg of voglibose daily to determine the effects of these drugs on UAE [92]. All these compounds significantly reduced HbA_1c after 3 months but only pioglitazone reduced UAE. Moreover, pioglitazone was associated with a decrease in office BP of 6/4 mmHg, whereas glibenclamide and voglibose produced slight increases of 2/1 and 4/2 mmHg, respectively. None of these changes was significant vs baseline but once again, no between‐groups comparison was reported. This is also the case for another study of the same group, in which 30 mg of pioglitazone was compared with placebo in 28 patients with type 2 DM and microalbuminuria [93]. After 6 months, apart from a significant reduction in UAE, pioglitazone produced an insignificant fall of 4 mmHg in office SBP. In the placebo group a rise of 6 mmHg in SBP was observed and again no between‐groups comparison to evaluate this 10 mmHg difference was performed. One study reporting no effect on BP with pioglitazone and documenting no BP change was that by Satoh et al. [94] (Table II), which aimed to elucidate the effect of the drug on CRP levels and pulse wave velocity in type 2 diabetic patients. In addition, Agarwal et al. [95] found no significant changes in ambulatory BP with pioglitazone, in a recent study aiming to compare the effects of pioglitazone and glipizide on proteinuria in subjects with overt diabetic nephropathy.

The positive effects of pioglitazone on BP are supported by the recently published PROactive (PROspective pioglitAzone Clinical Trial In macroVascular Events) Study, which was the first prospective outcome trial based on a TZD [96]. In this study, 5238 patients with type 2 diabetes and evidence of macrovascular disease were randomized to pioglitazone titrated from 15 to 45 mg or placebo in addition to pre‐existing glucose‐lowering drugs and other medications. After an average follow‐up of 34.5 months, pioglitazone was not superior to placebo as far as the primary endpoint was concerned (composite of all‐cause mortality, non‐fatal myocardial infarction, stroke, acute coronary syndrome, endovascular or surgical intervention in the coronary or leg arteries, and amputation above the ankle), but it was associated with a 16% lower risk for the main secondary endpoint (composite of all‐cause mortality, non‐fatal myocardial infarction, and stroke) [96]. Although BP changes was not a predefined endpoint, at the end of the study pioglitazone was also associated with a median change of −3/−2 mmHg in office BP levels, whereas the change in the placebo group was 0/−1 mmHg (p = 0.03 for SBP and p = 0.13 for DBP, respectively). It has to be noted that changes in antihypertensive drugs during the study were made in 1.4% of patients in the pioglitazone group and in 2.5% of those receiving placebo [96].

Studies with rosiglitazone

As with the two other TZD compounds, animal studies with rosiglitazone published until recently also reported a beneficial effect of this compound on BP levels. In these studies, rosiglitazone was shown to reduce BP levels or prevent development of hypertension in various animal models, such as fatty Zucker rats [97–99], Sprague–Dawley rats receiving angiotensin II [81], deoxycorticosterone acetate (DOCA)‐salt Sprague–Dawley rats [100], normal‐fed [101] and fructose‐fed [102,103] Sprague–Dawley rats, transgenic hypertensive mice expressing both human renin and human angiotensinogen transgenes [104] and SHR [105].

As with pioglitazone, human studies on the effect of rosiglitazone on BP changes were not available until a few of years ago. The first study reporting an effect of rosiglitazone on BP aimed primarily to investigate cardiac safety and the antihyperglycaemic effect of the drug in patients with type 2 DM [106]. A total of 203 subjects were randomized to 4 mg of rosiglitazone b.i.d. or glyburide, and after 52 weeks both drugs produced similar improvements in fasting plasma glucose and HbA_1c and neither adversely affected cardiac structure or function. Among 129 patients that performed ABPM, mean 24‐h DBP in the rosiglitazone group showed a significant reduction of 2.3±5.6 mmHg vs baseline, while 24‐h SBP did not change, as shown in Table III. Moreover, there was a significant difference in 24‐h ambulatory BP of −3.5/−2.7 mmHg in the rosiglitazone group compared with the glyburide group. However, no IS or plasma insulin measurements, neither correlations between BP and IS or insulin changes were reported. In the second relative study, Raji et al. [107], openly added rosiglitazone 4 mg b.i.d. to 24 non‐diabetic hypertensive subjects for 16 weeks and apart from a significant increase in IS of about 15% and a decrease in insulin levels of about 20%, they observed a significant reduction of 24‐h SBP (from 138±2 to 134±2 mmHg) and 24‐h DBP (from 85±2 to 80±2 mmHg). This decline in mean 24‐h SBP correlated significantly with the change in IS (r = 0.59, p<0.05) and showed a non‐significant correlation between the changes of mean 24‐h DBP and IS (r = 0.36, p<0.10).

Table III. Studies on subjects with components of the metabolic syndrome reporting effects of rosiglitazone on blood pressure levels.

Study	Type of subjects	n	Regimens compared	Duration	BP measurement	Mean effect on SBP/DBP vs baseline in Rosi groups (mmHg)	IS or IR measurement
St John Sutton M et al., Diabetes Care 2002 [106]	DM2, (hyp)^b	129	8 mg Rosi vs Gli	12 months	Ambulatory BP	−0.1^a/−2.3^a	–
Raji A et al., Diabetes Care 2003 [107]	hyp	24	8 mg Rosi	16 weeks	Ambulatory BP	−4^a/−5^a	e‐h clamp
Honisett S et al., Diabetes Care 2003 [109]	DM2, PW	31	4 mg Rosi vs plb	12 weeks	Office BP	−12^a/−6^a	–
Shargorodsky M et al., Am J Hypertens 2003 [108]	DM2, (hyp)^b	52	4 to 8 mg Rosi	6 months	Office BP	−20.1^a/−17.2^a	–
Wang TD et al., Am J Cardiol 2004 [110]	MS	50	4 mg Rosi vs plb	8 weeks	Office BP	−10^a/−7^a	–
Bennett SM et al., Diabet Med 2004 [111]	IGT, (hyp)^b	18	8 mg Rosi vs plb	12 weeks	Ambulatory BP	−7.0^a/−6.4^a	e‐h clamp
Natali A. et al., Diabetes Care 2004 [112]	DM2, (hyp)^b	74	8 mg Rosi vs 1500 mg Met vs plb	16 weeks	Ambulatory BP	−4/−2^a	e‐h clamp
Yosefy C et al., J Cardiovasc Pharmacol 2004 [113]	DM2, hyp	48	8 mg Rosi vs 10 mg Gli	8 weeks	Office BP	−6.1^a/−4.2^a	–
Sarafidis P et al., J Hypertens 2004 [114]	DM2, hyp	24	4 mg Rosi	6 months	Ambulatory BP	−5.4^a/−4.1^a	e‐h clamp HOMA‐IR
Kim YM et al., Diab Res Clin Pract 2005 [115]	DM2, (hyp)^b	125	4 mg Rosi vs no additional therapy	12 weeks	Office BP	−2.4/−2.9^a	HOMA‐IR, QUICKI
Negro R et al., Diab Res Clin Pract 2005 [116]	DM2, non‐dippers	38	8 mg Rosi vs plb	12 months	Ambulatory BP	Daytime: −3.4^a/−5.2^anight‐time: −5.1^a/−6.5^a	HOMA‐IR
Rosak C et al., Int J Clin Pract 2005 [117]	DM2, (hyp)^b	10,321	4‐8 mg Rosi	6 months	Office BP	−7^a/−3^a	–
Beck‐Nielsen H et al., Diabetologia 2005 [Abstract] [118]	DM2, (hyp)^b	668	Met+Rosi vs Met+Sulph and Sulph+Rosi vs Sulph+Met	12 months	Ambulatory BP	−4.8^a/−3.7^aand −3.8^a/−3.5^a	–
Ruilope LM et al., Diabetologia 2005 [Abstract] [119]	DM2, (hyp)^b	389	Met+Rosi vs Met+Gli	32 weeks	Ambulatory BP	−1.3^a/−2.3^a	–

SBP, systolic blood pressure; DBP, diastolic blood pressure; IS, insulin sensitivity; IR, insulin resistance; DM2, subjects with type 2 diabetes mellitus; IS, insulin sensitivity; IR, insulin resistance; hyp, subjects with hypertension; PW, postmenopausal women; MS, subjects with the metabolic syndrome; IGT, subjects with impaired glucose tolerance; Rosi, rosiglitazone; Gli, glibenclamide; plb, placebo; Met, metformin; Sulph, sulphonylurea; e‐h clamp, euglycaemic–hyperinsulinaemic clamp; HOMA‐IR, homeostasis model assessment – insulin resistance index; QUICKI, quantative insulin sensitivity check index. ^aSignificant change vs baseline levels or the other group compared. ^bA percentage of the total population was also hypertensive.

In another open study, rosiglitazone was added in 52 patients with type 2 DM treated with or without insulin, 58% of whom were also hypertensive for 6 months [108]. Rosiglitazone treatment was associated with significant reductions of HbA_1c and insulin levels, together with a rather substantial decrease in BP of 20.1/17.2 mmHg vs baseline, but ABPM or IS measurements were not carried out. Honisett et al. [109] added 4 mg of rosiglitazone or placebo on the background treatment of 31 postmenopausal type 2 diabetic, non‐hypertensive women. After 12 weeks of treatment, rosiglitazone was associated with significant improvement in glycaemic control, together with significant decreases in clinic SBP (from 124±10 to 112±9 mmHg) and DBP (from 71±4 to 65±3 mmHg) vs baseline but again no ABPM or IS measurements were performed. In a subsequent study, 50 non‐diabetic patients who met a modified National Cholesterol Education Program definition for the metabolic syndrome were randomized to either rosiglitazone (4 mg/day) or placebo [110]. After 8 weeks of treatment, office BP in the rosiglitazone group showed a significant reduction of 10/7 mmHg vs baseline.

Bennett et al. randomly assigned 18 subjects with persistent IGT to 4 mg of rosiglitazone twice daily or placebo [111]. At the end of the study fasting glucose values did not differ between the two groups, but the 2‐h plasma glucose levels following an oral glucose tolerance test (OGTT) were significantly lower and the IS index significantly higher with rosiglitazone. In addition, in the rosiglitazone group 24‐h BP showed a mean significant reduction of 7.0/6.4 mmHg vs baseline and a significant difference of –9.8/−8.0 mmHg compared with placebo. It should be noted, however, that mean baseline BP levels were much higher in the rosiglitazone group (132/76 mmHg) than in the placebo group (121/69 mmHg) and therefore the results of between‐groups comparison must be interpreted with caution.

In another interesting study, Natali et al. [112] randomly assigned 74 type 2 diabetic patients to rosiglitazone (8 mg/day), metformin (1500 mg/day), or placebo aiming to determine the effect of improving metabolic control and peripheral IR on vascular reactivity. After treatment for 16 weeks, rosiglitazone and metformin produced similar significant reductions in fasting glucose and HbA_1c levels, but IS increased only with rosiglitazone. This compound was also associated with an insignificant fall in 24‐h SBP of 4±2 mmHg, a significant fall in 24‐h DBP of 2±1 mmHg, as well as an improvement in the slope of the forearm blood flow response to acetylcholine, whereas no change in these parameters were observed with either metformin or placebo. Yosefy et al., after treating 48 type 2 diabetic, hypertensive and hyperlipidaemic subjects for 4 weeks with cilazapril and simvastatin, divided them into two groups to receive additional treatment of either 8 mg of rosiglitazone or 10 mg of glibenclamide for another 8 weeks [113]. Rosiglitazone was superior to glibenclamide in plasma glucose reduction and also produced a drop in plasma insulin levels. In addition, in the rosiglitazone group, a significant fall in office BP (6.1±4.1 mmHg for SBP and 4.2±1.9 mmHg for DBP) vs baseline levels was observed, whereas glibenclamide significantly increased SBP by 3.1±2.5 mmHg and produced no change in DBP levels.

In a pilot study from our group, 4 mg of rosiglitazone were added in the pre‐existing treatment of 24 patients with hypertension and type 2 DM for 26 weeks [114]. We originally aimed to elucidate the effect of this compound in patients for whom it was more often used in the clinical practice, who were at the same time in need for BP control. Therefore, all patients had type 2 diabetes without achieving full glycaemic control although receiving the maximum daily dose of glibenclamide. All were hypertensives without adequate BP control. Moreover, we focused on the determination of the possible association between IS and BP changes, and for this reason both clamp and ABPM measurements were included, something that was previously done in only a few relevant studies [107],[111,112]. At the end of the treatment, rosiglitazone produced a significant increase in clamp‐estimated IS of about 25%, and a corresponding reduction of the HOMA‐IR index, as well as significant decreases in fasting glucose, fasting insulin and HbA_1c levels. In addition, it was associated with significant reductions of 5.4 and 4.1 mmHg in mean ambulatory SBP and DBP respectively, as well as significant decreases of pulse pressure, systolic and diastolic load. The changes in SBP and DBP levels recorded were strongly correlated with the changes of IS (r = −0.78 for SBP, and r = −0.68 for DBP), ΗOMA‐IR index (r = 0.44 and r = 0.45 respectively) and fasting plasma insulin levels (r = 0.48 and r = 0.46 respectively) [114].

In addition, Kim et al. randomized 125 Korean patients with type 2 DM to treatment with 4 mg of rosiglitazone daily or to a control group for 12 weeks to evaluate the efficacy of the drug. Rosiglitazone was again associated with significant reductions of fasting glucose and HbA_1c levels, as well as a significant decrease in HOMA‐IR. Office SBP decreased 2.4 mmHg, which was not significant, but DBP showed a significant fall of 2.9 mmHg vs baseline [115]. In a more recent study from Negro et al., 38 non‐dipper type 2 diabetic patients already on metformin were randomly treated with 8 mg of rosiglitazone or placebo for 12 months [116]. In contrast to their previous study of short duration [88], the authors observed significant decreases in fasting glucose, insulin, HbA_1c and HOMA‐IR index. In addition, by using ABPM they were also able to document small but significant reductions in BP levels. At the end of the study daytime ambulatory BP showed a significant drop of 3.4/5.2 mmHg and night‐time BP a significant drop of 5.1/6.5 mmHg vs baseline. All these BP changes were significantly correlated with the reduction of HOMA‐IR index [116].

Rosak et al. presented data from a total of 11,014 type 2 diabetic subjects from two large cohort observational studies who were treated with rosiglitazone in addition to metformin for 6 months [117]. In the 10,321 subjects who completed the final evaluation, the addition of rosiglitazone was associated with a significant reduction in median HbA_1c by 1.3% and fasting plasma glucose by 47 mg/dl. In addition, SBP fell from 144±15 at baseline to 137±12 mmHg at the end of the study and DBP from 85±9 to 82±7 mmHg, which was a significant reduction of 7/3 mmHg (p<0.0001). These findings are certainly limited by the observational nature of the study and the fact that the investigator was able to adjust the dose of any anti‐hypertensive medication during the treatment period. However, it is worth mentioning that data from the larger of the two pooled observational studies (representing about two‐thirds of the subject population) showed that the proportion of subjects receiving anti‐hypertensive medication and the distribution of subjects receiving mono‐ or combination anti‐hypertensive therapy remained almost unchanged.

This beneficial effect of rosiglitazone on BP is strongly supported by the recently presented results of two prospective, double‐blind, active treatment‐controlled studies that used ABPM, included a large number of patients and lasted more than 6 months. The first of them was a sub‐study of the Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of glycaemia in Diabetes (RECORD) trial, aiming to determine the effect of this compound on BP levels [118]. With this aim, a sub‐group of 759 patients with type 2 diabetes, inadequately controlled by metformin or sulphonylurea were randomized to add‐on rosiglitazone vs sulphonylurea or rosiglitazone vs metformin respectively and had a 24‐h ABPM at baseline, 6 and 12 months at follow‐up. In 668 subjects with analysable ABPM data at 12 months available, the addition of rosiglitazone to metformin produced greater reductions in ambulatory SBP and DBP than the addition of sulphonylurea (−4.8 vs −2.1 mmHg, p = 0.009 and −3.7 vs −1.7 mmHg, p = 0.002 respectively). Moreover, the addition of rosiglitazone to sulphonylurea was associated with greater reductions in ambulatory SBP and DBP than the addition of metformin (−3.8 vs −1.1 mmHg, p = 0.02 and −3.5 vs −0.4 mmHg, p<0.0001 respectively). In the study subjects with a previous diagnosis of hypertension, the addition of rosiglitazone also led to significant greater reductions of BP levels. Changes in the use of various antihypertensive medications were similar across all treatment groups. In the second study, Ruilope et al. [119] randomized 389 patients with type 2 diabetes and microalbuminuria receiving metformin to rosiglitazone or glibenclamide. The primary aim of the study was to evaluate the effect of rosiglitazone on UAE, but 24‐h ABPM was also performed to assess changes in BP levels. After 32 weeks of treatment, in parallel to a decrease in UAE of about 23%, the addition of rosiglitazone was associated with a mean reduction in SBP of 1.3 mmHg vs baseline, whereas the addition of glibenclamide with an increase of 1.3 mmHg (p = 0.01). In addition, rosiglitazone lowered DBP by 2.3 and glibenclamide increased it by 0.3 mmHg (p<0.01).

Data from studies in special populations

The effects of TZDs on BP have been evaluated not only in the typical subjects with various components of the metabolic syndrome, but also in patients suffering from other conditions. For example, two studies have examined the effect of TZDs on BP in patients with end‐stage renal disease. In the first of these studies rosiglitazone or pioglitazone were added in 40 patients with diabetes receiving hemodialysis treatment for a period of 3 months [120]. Combined thiazolidinedione data yielded insignificant effects for all clinical and laboratory parameters measured, with the exception of HbA_1c and BP measured before dialysis (−5.57±12.09 mmHg for SBP and −3.24±6.17 mmHg for DBP). Lin et al. studied the effects of 4 mg of rosiglitazone in 15 non‐diabetic patients on continuous ambulatory peritoneal dialysis therapy. After 3 months of treatment rosiglitazone had significantly improved insulin sensitivity but it was not associated with significant change in HbA_1c or office BP levels [121].

The effect of rosiglitazone on BP was also investigated in non‐obese women with polycystic ovary syndrome (PCOS) whose insulin sensitivity was normal [122]. In that study, 100 patients with PCOS randomly received 850 mg of metformin, 4 mg rosiglitazone, a combination of both treatments or placebo twice a day for 6 months. At the end of the study, apart from an increase in the frequency of ovulatory circles, office BP levels were significantly decreased in all actively treated groups (change from baseline −4.4/−1.5, −2.4/−2.0, and −4.5/−2.1 mmHg for metformin, rosiglitazone and the combination group, respectively).

In addition to the studies above, the use of TZDs have been evaluated in subjects with Human Immunodeficiency Virus (HIV) infection, which after receiving antiretroviral therapy commonly develop lipoatrophy and insulin resistance. In particular, Kovacic et al. randomized HIV‐infected, lipoatrophic adults receiving antiretroviral therapy to 4 mg of rosiglitazone twice daily or matched placebo. Patients treated with rosiglitazone experienced significant reductions in insulin, IR and office SBP levels (−8 mmHg) in comparison to placebo‐treated patients after a treatment period of 48 weeks [123]. In a study from Gavrila et al., 14 patients with highly active antiretroviral therapy (HAART)‐induced metabolic syndrome were randomly assigned to pioglitazone, fenofibrate or matched placebo in a 2×2 factorial design. After 12 months of treatment, pioglitazone was found to improve IR and fasting insulin levels and to produce a significant decrease of about 14.1 mmHg in office SBP and an almost significant (p = 0.06) decrease of 5.4 mmHg in DBP, whereas fenofibrate did not have any effect on these parameters [124].

In another very interesting recent study, Strowig & Raskin evaluated the efficacy of rosiglitazone in the treatment of overweight subjects with type 1 diabetes, many of them with elevated IR [125]. Fifty adult patients with type 1 diabetes were randomized in insulin and placebo or insulin and 4 mg of rosiglitazone twice daily, with insulin regimen modified in all subjects to achieve near‐normal glycaemic control. After 8 months of treatment, both groups showed a significant reduction of HbA_1c levels, as well as significant weight gain of comparable magnitude, but only in the group on rosiglitazone office BP improved significantly vs baseline (from 137.4±15.6 to 128.8±14, and from 87.2±9.4 to 79.4±7.2 mmHg for SBP and DBP respectively).

Conclusions

For almost two decades, a number of mechanisms linking IR and compensatory hyperinsulinaemia with BP elevation have been elucidated or at least hypothesized, strengthening the presence of hypertension among the disorders clustering in the Metabolic Syndrome. The introduction of TZDs, a class of antidiabetic compounds that act through IR reduction, has provided new perspectives, since lowering of IR could also be beneficial for other aspects of the syndrome, according to the hypotheses for its aetiology. An important number of studies have therefore tried to investigate this possibility, several of which focused mainly on BP changes. Data from animal studies was encouraging, since all TZDs have been found to lower BP or attenuate the development of hypertension in a variety of animal models. The majority of human studies in this field also show a beneficial impact of all TZDs on BP, as extensively discussed in the present review. In some of these studies, the change in BP was correlated with IR and/or plasma insulin changes, findings in favour of the importance of IR and chronic hyperinsulinaemia reversion in BP amelioration.

Although human studies on the effect of TZDs on BP have increased in number in the last few years and provided tantalizing data from a variety of subject categories (from subjects with or without type 2 diabetes, IGT, hypertension, obesity, etc. to patients with end‐stage renal disease and HIV infection), the definite answer did not seem to be reached. Most of these studies did not use the most adequate method for BP measurement, whereas other studies were small in size, short in duration or not properly designed. Therefore, long‐term, prospective, randomized, placebo‐controlled trials with one of the TZDs in clinical use, including a substantial number of patients and a BP measurement method capable to elucidate small differences were needed to clarify this issue. To this direction, newly presented data seem to provide important evidence in favour of this modest but significant effect of TZDs on BP [118,119]. It has also to be noted that a possibility of publication bias, with studies showing large and significant effects of glitazones on BP being more often published than others with smaller or no effects, should always be taken into consideration. However, the appearance of the respective funnel plots presented in Figure 1, which, with the exception of one study [108], seem fairly symmetrical, and the fact that most of the studies in this report did not include BP changes as a primary end‐point suggest that publication bias related to BP changes by glitazone therapy is less likely. Overall, given the great cardiovascular benefit that patients with or without type 2 diabetes can have from modest or even minor decrements in BP levels [126,127], if this effect of TZD would be eventually confirmed, it could be valuable within the framework of a multifactorial intervention.

Figure 1. Funnel plots illustrating the relations between the effect of glitazones on systolic (A) and diastolic (B) blood pressure and the size of the glitazone group in each study. The plots include all reviewed studies in patients with components of the metabolic syndrome, with the exception of the papers of Belcher et al. [89] and Rosak et al. [117], which reported combined data from more than one original study, and the PROACTIVE study [96], where alterations in background antihypertensive treatment were permitted and blood pressure changes are reported as median values (SBP, systolic blood pressure; DBP, diastolic blood pressure; •, studies with troglitazone; □, studies with pioglitazone; ◊, studies with rosiglitazone). Weighted mean changes in the larger study (118) indicated by (‐‐‐‐‐).